Journal of Petrology Advance Access originally published online on May 6, 2005
Journal of Petrology 2005 46(9):1901-1924; doi:10.1093/petrology/egi042
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Garnetite Xenoliths and Mantle–Water Interactions Below the Colorado Plateau, Southwestern United States
1 DEPARTMENT OF GEOLOGICAL SCIENCES, JOHN A. AND KATHERINE G. JACKSON SCHOOL OF GEOSCIENCES, THE UNIVERSITY OF TEXAS AT AUSTIN, 1 UNIVERSITY STATION C-1100, AUSTIN, TX 78712-0254, USA
2 GEMOC KEY CENTRE, DEPARTMENT OF EARTH AND PLANETARY SCIENCES, MACQUARIE UNIVERSITY, SYDNEY, NSW 2109, AUSTRALIA
3 CSIRO EXPLORATION AND MINING, NORTH RYDE, NSW 2113, AUSTRALIA
RECEIVED JULY 7, 2004; ACCEPTED MARCH 21, 2005
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONGarnetite xenoliths from ultramafic diatremes in northeastern Arizona provide insights into hydration and metasomatism in the mantle. The garnetites contain more than 95% garnet, some of which has complex compositional zonation related to growth in fractures within grains. Accessory minerals include rutile, ilmenite, chlorite, clinopyroxene, and zircon. Zircon grains in one rock were analyzed in situ to determine U–Pb ages and Hf isotopic compositions. Most U–Pb analyses plot on or near concordia in the range 60–85 Ma but a few are discordant. The range in 176Hf/177Hf is about 0·2818–0·2828, with grains zoned to more radiogenic Hf from interiors to rims. The garnetite protolith contained zircons at least 1·8 Ga in age, and garnet and additional zircon crystallized episodically during the interval 85–60 Ma. The garnetites are interpreted as mantle analogues of rodingites, formed in metasomatic reaction zones caused by water–rock interactions in Proterozoic mantle during late Cretaceous and Cenozoic subduction of the Farallon plate. Associated eclogite xenoliths may have been parts of these same reaction zones. The rodingite hypothesis requires serpentinization in the mantle wedge 700 km from the trench, beginning 5–10 Myr before tectonism related to low-angle subduction.
KEY WORDS: garnetite; Lu–Hf, mantle; rodingite; metasomatism
|
INTRODUCTION |
|---|
Garnetite xenoliths from a diatreme cluster in the Navajo volcanic field of the Colorado Plateau have been studied to investigate implications for serpentinization and metasomatism of continental mantle. Helmstaedt & Schulze (1988)
suggested that these xenoliths formed by reactions involving peridotite hydration. If so, they may be samples of metasomatic reaction zones analogous to those that form at low temperatures and pressures during serpentinization (Coleman, 1967); garnet-rich parts of such zones are called rodingites. If the garnetites are analogues of rodingites, then they may provide evidence of mantle processes associated with subduction of the Farallon plate, by analogy with origins proposed for eclogite xenoliths from the same diatremes. The locality is about 700 km from the Farallon subduction zone according to the reconstruction of Severinghaus & Atwater (1990); however, this is more than three times the maximum distance from the trench that Hyndman & Peacock (2003) considered a typical limit for subduction-induced serpentinization. Evidence that the garnetites are rodingite analogues would be consistent with the hypothesis of Humphreys et al. (2003) that low-angle subduction hydrated the mantle wedge below a broad region of western North America.
The garnetites also have been studied to clarify the genesis of the associated eclogite xenoliths. The eclogites have been interpreted either as fragments of the Farallon slab itself (Helmstaedt & Doig, 1975; Usui et al., 2003) or as products of water–rock reactions in the mantle wedge (Smith et al., 2004). These contrasting hypotheses imply very different effects of low-angle subduction on the mantle roots of continents. The garnetite xenoliths are hosted only by those diatremes that host the eclogites, and Switzer (1975) suggested that these rock types were genetically related. If so, the garnetites yield insights into how the eclogites formed.
|
OCCURRENCE |
|---|
All garnetite xenoliths studied here are from the Garnet Ridge diatreme cluster on the Comb Ridge monocline in northeastern Arizona (Fig. 1). The host rocks in the diatremes are serpentinized ultramafic microbreccia (SUM; Roden, 1981
). These SUM host rocks appear to have been emplaced as gas–solid mixes (McGetchin & Silver, 1972), and textural and chemical evidence supports the conclusion that a melt phase was never part of the mixture (Roden, 1981). Rather, the gas–solid eruptions have been interpreted as products of hydrated mantle disaggregated during heating by intruded magma (Smith & Levy, 1976). The Garnet Ridge diatremes were emplaced at 30 Ma (Smith et al., 2004).x
|
, location of the Garnet Ridge and other major diatremes of serpentinized ultramafic microbreccia (SUM) in the Navajo volcanic field. The boundary of the Navajo field is determined by the distribution of minette and related rocks.Garnetite occurs in the diatremes together with an extraordinary variety of xenoliths of sedimentary, metamorphic, and igneous rocks: gabbro, granite, rhyolite, granulite, amphibolite, peridotite, omphacite pyroxenite, and eclogite are among the rock types present. Eclogite makes up less than 0·05 wt % of the population of igneous plus metamorphic rocks (Hunter, 1979
). The abundance of garnetite xenoliths was not measured, but they are less common than those of eclogite, and they probably make up less than 0·01 wt % of that population. Garnetite also occurs in the two other major diatremes on the Comb Ridge monocline, Mule Ear and Moses Rock (Helmstaedt & Schulze, 1988; McGetchin & Silver, 1970). |
DESCRIPTIONS AND PETROGRAPHY |
|---|
The largest of the 14 garnetite xenoliths studied in thin section had a maximum diameter of 8 cm, slightly smaller than the 11 cm dimension of the largest garnetite xenolith described from the province by Switzer (1975)
. Most were less than 4 cm in maximum diameter. Typical specimens are nearly equant and have smooth surfaces, some of which appear polished. These smooth surfaces are attributed to abrasion and impact during emplacement of the gas–solid mix that formed the diatreme fill, as described by McGetchin & Silver (1970, 1972). Irregular fractures, some coated with films of secondary calcite, perhaps caliche, are present in most samples. The garnetite xenoliths consist of 95% to almost 100% garnet. Grain boundaries are difficult to identify. Slight variations of garnet color from clear to very pale orange are visible in sections of some rocks. These color variations were seen only in polished thin sections with thicknesses in the range from 100 to 300 µm. In some rocks, parts of garnets are turbid with unoriented inclusions a few micrometers in average diameter, many of which appear to be of fluid. The grain boundaries of garnets were inferred primarily from the color variations and turbidity and from cracks. Where estimates were possible, garnet grains appeared anhedral and approximately equant, typically with maximum diameters of a few millimeters.
A diverse suite of minor and trace minerals is present; multiple polished thin sections were prepared of most samples so as to identify as many as possible. In approximate order of decreasing abundance, these minerals are: rutile, ilmenite, chlorite, clinopyroxene, zircon, pyrite, phlogopite, and apatite. Rutile and ilmenite are the only minerals other than garnet present at a modal abundance of 1% or more; the others are present only as trace constituents. Grains of the two oxides typically are anhedral, with maximum dimensions of several millimeters or less. Rutile forms an estimated 5% of one sample, but it makes up less than 2% of more typical samples and is absent in some. The maximum abundance of ilmenite was estimated as about 2%; most samples contain <<1%, and it was not observed in some rocks. Chlorite was identified in six of the 14 samples, as was clinopyroxene. Apatite, phlogopite, and pyrite were identified in two to four rocks. Clinopyroxene, chlorite, and apatite are slightly more common at the edges of several of the xenoliths in which they occur. In rocks that contain both ilmenite and rutile, they occur as discrete grains and as rims around one another. Chlorite is intergrown with rutile in some occurrences and with ilmenite in some others. No other systematic mineral associations were noticed.
Zircon was identified in five xenoliths, but only in one of these rocks (GR1-201) were more than a few grains observed. Although zircon is present only in trace proportion, it is the most common silicate other than garnet in that rock. Zircon grains in GR1-201 were chosen for in situ U–Pb and Lu–Hf analysis. Most grains are anhedral to subhedral, but rare grains are euhedral; maximum diameters range from several micrometers to about 300 µm.
|
MINERAL CHEMISTRY |
|---|
Procedures for elemental analysis
Minerals in 11 xenoliths were analyzed by wavelength-dispersive spectrometry (WDS) on JEOL 7300 and 8200 electron microprobes at The University of Texas at Austin. For most analyses, the accelerating voltage was 15 keV and the beam current was about 40 nA. Counting at peak and background wavelengths was terminated at 40 s for each, unless a standard deviation <0·3% was achieved first, based on counting statistics. In a few analytical sessions, trace elements were analyzed with a 20 keV accelerating voltage, higher beam current, and longer times for data acquisition. Data were corrected with a JEOL ZAF procedure. Representative electron microprobe (EMP) analyses of minerals in six of the xenoliths are given in Table 1, and representative analyses for all of the xenoliths are given in the Electronic Appendix (available at http://www.petrology.oupjournals.org).
|
Trace elements in minerals of five rocks were analyzed by laser ablation microprobe–inductively coupled plasma mass spectroscopy (LAM–ICPMS) at The University of Texas. After photography and characterization by EMP analysis, minerals in polished thin sections were ablated with a New Wave LUV213 laser. The Nd:YAG source delivers a focused 213 nm beam; resulting ablation pits have diameters from 50 to 100 µm. The ablated material was transported with a helium carrier gas into a Micromass Platform quadrupole inductively coupled plasma mass spectrometer. Signal intensities were tabulated for each mass sweep in most of the analyses, and anomalous data were excluded. Concentrations in garnet and clinopyroxene were calculated with 44Ca as an internal standard in two of the three datasets and with 43Ca in the other: reanalysis verified consistency of the two choices. Two semiquantitative analyses of rutile were calculated with 93Nb as an internal standard. Four NIST SRM glasses (610, 612, 614, and 616) and USGS glass BCR-2G were analyzed during each session. For almost all analyzed masses, calibration curves were fitted either to SRM 616, 614, and 612 or to all four of the NIST glasses, depending upon the relative count rate for the unknown. Glass BCR-2G was used as a secondary standard, except in rare cases when it was included as a primary standard to avoid extrapolations to much higher concentrations than those in NIST SRM 610. Consistency of some of the garnet and pyroxene analyses was tested and verified by analysis of two masses for each of five REE. Accuracy was verified by the analyses of the secondary standard, BCR-2G. Backgrounds were obtained by measurements without ablation of samples. Detection limits were not calculated from these backgrounds, because of possible differences between signal levels with and without sample ablation. Instead, results are not reported below determination limits based on average concentrations within two standard deviations of zero. Representative determination limits were about 55 ppb for La in garnet and near and below 12 ppb for Tm, Yb, and Lu in pyroxene.
Garnet
The most common garnet compositions have roughly equal atomic proportions of Mg, Fe, and Ca (Fig. 2), consistent with the more limited data of McGetchin & Silver (1970), Switzer (1975) and Helmstaedt & Schulze (1988). The compositions average about 2% andradite, calculated from (2 – Al – Cr)/2 and 8-cation formulae. Other minor elements are present with the following average contents: TiO2, 0·07 wt %; Cr2O3, 0·05 wt %; MnO, 0·34 wt %; Na2O, 0·03 wt %. No clear correlations were observed between major (Ca, Fe, Mg) and minor elements (Na, Ti, Cr, and Mn) in the entire database, although Fe and Mn are well correlated in the garnet of some rocks. The most magnesian garnets (50% pyrope end-member) are found in xenolith N379b-GR-4, and they have more Cr2O3 (0·13–0·17 wt %) than garnet in other samples.
|
Many garnets are complexly zoned, as illustrated in back-scattered electron (BSE) images in Fig. 3. Two types of zoning were distinguished. First, gradients from grain interiors to rims were observed (Fig. 3a and b). Second, compositional gradients also are present in patterns that record both growth of garnet within fractures and reaction of the fractured grains (Fig. 3c). The more complex patterns formed by the interplay of both processes (Fig. 3d and e).
|
No trends of compositional zoning are consistent for the full dataset, although trends in individual garnets are well defined (Figs 4 and 5). BSE images document compositional changes at rims adjacent to ilmenite in rock N375-GR (Fig. 3a and b): one such rim, about 400 µm wide, is Mg-rich and Ti-poor relative to the interior (Fig. 4a). Annular orange zones are present in garnets of rock GR1-202, and a traverse across such a zone and an apparent grain boundary is plotted in Fig. 4b; Ti is high at the boundary relative to the grain interior, unlike the example plotted in Fig. 4a. The gradients in the garnet of Fig. 5 are due to both ‘normal’ and fracture-related growth. In a clear case of fracture-related zonation in N379b-GR-2, the fracture fill is less calcic and less titaniferous than the surrounding garnet (Fig. 5a). Garnet in N379b-GR-1 has complex zoning that includes both interior-to-rim gradients and gradients along sets of small intersecting fractures (Fig. 5b).
|
error bars show the range of 4 SD for a representative analysis, calculated solely on the basis of counting statistics. (a) Apparent growth zoning in garnet at a contact with ilmenite in rock N375-GR, along a path marked on the BSE image in
|
Trace elements determined by LAM–ICPMS in garnet include Ni, Y, Zr, Yb, and Hf and all the rare earth elements (REE) except La (Table 2). Measurements of La, Nb, Ta, and U yielded concentrations below determination limits. Average Ni concentrations fall in the range 8–18 ppm. Of garnet in the five xenoliths, that in zircon-bearing GR1-201 is highest in Y (67 ppm), Lu (1·3 ppm) and the other heavy REE (HREE), and lowest in Zr (2 ppm) and Hf (0·04 ppm). Chondrite-normalized profiles of REE abundance in garnet in each of the five rocks are relatively similar for the range Ce to Sm (Fig. 6a). In three of the rocks, however, normalized REE abundances decrease with increasing atomic number in the range Sm to Lu, forming ‘humped’ patterns. In one of these three rocks, there is a distinct positive Eu anomaly, and in the other two, poorly defined negative anomalies. The patterns of average REE abundance for each rock adequately represent all the analyses of garnet in that sample, except for garnetite N379b-GR-1. Analyses of five of the six ablated volumes of garnet in this rock form a clustered group, whereas the sixth is much lower in HREE (Fig. 6b). The volume of this anomalous analysis is distinguished by lower average atomic weight and so is relatively dark on the BSE image (Fig. 3d); the volume also has relatively lower Fe and higher Mg, Ca, and Cr (Fig. 5b). Although no comparable inhomogeneity in REE was documented in garnet of the other four samples, analytically significant differences in concentration are present (Fig. 6b and c).
|
|
Rutile, ilmenite, clinopyroxene, chlorite, and phlogopite
Compositions of rutile and ilmenite are similar in all but xenolith N379b-GR-4 (Table 1 and the Electronic Appendix). Ilmenite, analyzed by EMP in five xenoliths, contains 0·09–0·27 wt % NiO and 4·0–7·1 wt % MgO and trace Cr, Al, and Mn; that in sample N379b-GR-4 has the highest Mg, Ni, and Cr. EMP analyses of rutile in seven xenoliths establish the typical range of 0·1–0·3 wt % Cr2O3, but rutile in N379b-GR-4 contains 0·5–0·6 wt % Cr2O3. In two of the three rocks in which rutile was analyzed for Nb2O5, concentrations are near 0·04 wt %, but they are much higher, 0·62–1·20 wt % Nb2O5, in N379b-GR-4. Concentrations of V and Zr in rutile, determined by semiquantitative LAM–ICPMS analysis, are about 4500 and 170 ppm, respectively, in xenolith N379b-GR-4 and about 3000 and 60 ppm in the relatively zircon-rich rock, GR1-201.
Clinopyroxene was analyzed in five xenoliths (Tables 1 and 2, and Electronic Appendix). In three of the five, only one small grain was found, and in the fourth, only two grains. In three xenoliths, the Jd component of the pyroxene is in the range 3–6%, and in one it is 14%; in the xenolith with two identified grains, one has 12% Jd and the other 25%. Aegirine, calculated as (Na – Al – Cr) on a four-cation basis, ranges from 0·5% to 6%. The dominant end-member is diopside, and calculated Fe2+/(Fe2+ + Mg) of these pyroxenes ranges from 0·05 to 0·08. A relatively large grain of clinopyroxene, about 3 mm in maximum diameter, was found in one rock (ATG-GRP1-B), together with four much smaller grains. The chondrite-normalized pattern of REE abundances in the large grain has a maximum at Nd and very low values for the HREE (Fig. 6c). Concentrations of Ni and Sr are about 400 ppm; those of other trace elements are much lower.
Chlorite, analyzed by EMP in five xenoliths, is clinochlore, with Fe/(Fe + Mg) in the range 0·05–0·08 (Table 1 and Electronic Appendix). NiO contents range from 0·24 to 0·44 wt % and average 0·34 wt %. Fe/Mg of chlorite inclusions varies systematically with that of the surrounding garnet (Fig. 7). Phlogopite was identified in two rocks and analyzed in one. The phlogopite has Fe/(Fe + Mg) = 0·10 and 0·22 wt % NiO.
|
|
ZIRCON GEOCHRONOLOGY |
|---|
Analytical techniques
Zircons in four polished sections of garnetite GR1-201 were analyzed at GEMOC in Macquarie University. U–Pb ages were determined by LAM–ICPMS as described by Belousova et al. (2001)
and Jackson et al. (2004). The instrumentation consisted of a modified Merchantek/New Wave LUV 213 nm Nd:YAG laser attached to an Agilent 4500 s ICPMS system. Typical ablation pits are 30 µm in diameter, and all ablations were performed in a He carrier gas. Mass bias and instrumental drift were corrected using a very homogeneous external standard, the GEMOC GJ-1 zircon (608 Ma). Data were reduced using the in-house GLITTER software (www.es.mq.edu.au/GEMOC), which allows online selection of the most stable part of the time-resolved signal. U and Th contents of the ablated volumes were estimated by comparison of count rates with those of the GJ-1 standard, and probably are accurate to ±10%.
Hafnium isotopes were analyzed with a Merchantek/New Wave LUV 213 nm LAM, attached to a NuPlasma multicollector (MC) ICPMS system; ablations were performed in He. Typical ablation pits are 50–60 µm in diameter. Procedures for the correction of isotopic interferences of 176Lu and 176Yb on 176Hf and for mass bias corrections have been described by Griffin et al. (2000). Interpretations of Hf isotope evolution utilize the 176Lu decay constant of Blichert-Toft et al. (1997) and the model for the depleted-mantle source adopted by Griffin et al. (2000). U–Pb and Hf isotope data on the secondary standards analyzed as unknowns with each run are summarized in Table 3.
|
U–Pb zircon ages
All but four of the 27 analyses of 206Pb/238U and 207Pb/235U in zircons yield ages concordant within analytical uncertainties in the range from about 85 to 60 Ma (Fig. 8, Table 4). The 206Pb/238U ages cluster in two ranges, the larger cluster centered at about 70 Ma, and a smaller one centered at about 85 Ma (Fig. 8b). These U–Pb analyses are compatible with a model of episodic zircon growth extending over the period from about 85 to about 60 Ma. The most discordant point has a 207Pb/206Pb age of about 670 Ma. The discordant points document inheritance of a relatively small mass of zircon of Proterozoic but otherwise poorly defined age.
|
|
The analyzed zircons show a wide range in their contents of Th (3–237 ppm, mean 87 ppm) and U (35–436 ppm, mean 162 ppm). About one-third of the grains have Th/U
0·3 (median 0·49), which might be expected in metamorphic zircons, but others range up to >2. There is no correlation between age and Th/U, but the two oldest grains also have the highest Th contents and the highest Th/U. These older ages may reflect incomplete Pb loss from originally magmatic grains inherited from the protolith of the garnetite. The high mean contents of U and Th are unusual in metamorphic zircons.
Hf isotope ratios and age implications
The analyzed zircons have extraordinarily low 176Lu/177Hf and 176Yb/177Hf ratios, comparable with or lower than those found in kimberlitic zircons (Griffin et al., 2000). The depletion of the HREE is consistent with growth in the garnetite matrix, where HREE are strongly partitioned into garnet, but Hf partitions into zircon. There is no correlation between 176Hf/177Hf and Yb/Hf or Lu/Hf ratios, confirming the efficacy of the isobaric-overlap corrections employed here and described by Griffin et al. (2000).
The 176Hf/177Hf determinations extend over the range 0·281829(9)–0·282851(10) (Figs 9 and 10, Table 5). Hf isotope ratios in interiors and rims of selected grains were obtained both by multiple analyses and by ablating through zircons. There is a crude correlation of 176Hf/177Hf with percent HfO2; the zircons with least radiogenic Hf values have a median of about 2·1 wt % HfO2, and the most radiogenic a median of about 1·7 wt %. Also, there is a crude correlation of more radiogenic Hf with younger 206Pb/238U ages (Fig. 9b). Furthermore, in all cases, grain rims have higher 176Hf/177Hf than interiors, and a large part of the total range is present within single zircon grains (Fig. 10). Some of the grains were characterized by BSE and cathodoluminescence (CL) before analysis, and representative images with positions of ablated pits are shown in Fig. 11. Most of the analyzed grains are relatively large and have smooth shapes (Fig. 11a–c); the interiors of these grains have relatively old U–Pb ages and relatively low 176Hf/177Hf, consistent with the presence of an inherited Proterozoic component in the lased volumes. Zircons are also present that have botryoidal shapes and granular surfaces (Fig. 11d and e). Only two grains with this unusual morphology were large enough to analyze, and the ablated volumes in each had relatively high 176Hf/177Hf (
0·2825). Zircons with similar granular surfaces were noted in two of the fractions from Navajo eclogites used for U–Pb geochronology by Smith et al. (2004): neither of those multigrain fractions recorded inheritance of Proterozoic lead, in contrast to many of the other analyzed fractions from the eclogites. The botryoidal morphology may characterize some of the zircon formed in the Cenozoic metamorphic events.
|
|
|
|
The least radiogenic 176Hf/177Hf values yield depleted-mantle model ages of about 1·9 Ga (Fig. 9), and the age of the garnetite protolith is likely to have been at least 1·8 Ga. The depleted-mantle line in Fig. 9 reproduces average mid-ocean ridge basalt (MORB) values (Griffin et al., 2000
): the range for basalt on the East Pacific Ridge (Chauvel & Blichert-Toft, 2001), plausibly from mantle sources like those for Farallon MORB, is indicated by the EPR bracket in Fig. 9b. All values of 176Hf/177Hf in the zircons are less radiogenic than those expected for the Mesozoic and Cenozoic basalts of the Farallon plate. The more radiogenic 176Hf/177Hf values can be produced by episodic zircon growth in an environment with high Lu/Hf. The eight analyses of garnet in sample GR1-201 have a median Lu/Hf equal to 33, far higher than the values of 1–2 characteristic of garnet in the other garnetites. The calculated evolution of 176Hf/177Hf in an environment with Lu/Hf of 30 and with an initial value like that of the inherited zircon nicely fits the observed correlation with 206Pb/238U ages during the period from 85 to 70 Ma (Fig. 9b). Hf-isotope composition has a weak negative correlation with U: the grains with the most radiogenic Hf (lowest TDM, where DM indicates depleted mantle) have the lowest U contents. Most low-Th grains also have low TDM. There is considerable scatter in Th and U contents, however, and no correlation was recognized between Hf isotope composition and Th/U.
|
PETROLOGICAL FRAMEWORK AND CONSTRAINTS |
|---|
Processes of formation of other garnetites
Garnetites, rocks that consist mostly of garnet, have been attributed to a variety of processes. For instance, monomineralic garnet xenoliths in African kimberlite have been interpreted as magmatic cumulates; the garnets contain about 70% pyrope and 10–20% grossular end-members (Exley et al., 1983
). Layers of garnetite consisting of 60–90 vol. % garnet occur in high-grade amphibolite- and granulite-facies rocks in the Alps, and they have been attributed to anatexis and melt–rock interactions (Rivalenti et al., 1997); garnets in the most garnet-rich layers contain about 50% almandine and 10% grossular end-members. Garnetite with about 20 vol. % corundum occurs in a garnet peridotite lens in the Sulu ultrahigh-pressure terrane and has been interpreted as a metamorphic rock derived from a protolith of spinel websterite (Zhang et al., 2004). Grossular-rich garnetite boudins in migmatite have been attributed to ultrahigh-pressure metamorphism of a subducted calc-silicate protolith (Vrana & Fryda, 2003).
The most analogous garnet-rich rocks may be rodingites. Rodingites are formed in metasomatic reaction zones during serpentinization of peridotite in low-temperature near-surface environments, and they occur near contacts between peridotite and a variety of other lithologies (Coleman, 1967; Leach & Rodgers, 1978; Schandl et al., 1989; O'Hanley et al., 1992; Dubinska et al., 2004). Lenses of rodingite typically are less than 30 cm wide and are rarely more than 5 m in maximum dimension. The most common garnet in rodingite is fine-grained hydrogrossular; andradite is much less common. Most rodingites contain other minerals, such as diopside, prehnite, zoisite, and tremolite. Temperatures calculated from the mineral assemblages typically are in the range 200–500°C, but the higher temperatures may record metamorphism after rodingite formation, as described by Rice (1983). Some instances of the recorded metamorphism occurred at crustal pressures (e.g. Frost, 1975) and others at mantle pressures (e.g. Evans et al., 1979). A metarodingite–eclogite suite at Cima di Gagnone in the Swiss Alps may be of particular pertinence to the Navajo garnetites, because these Alpine rocks have been interpreted to record maximum pressures near 2·5 GPa (Evans et al., 1979, 1981). Garnets in the most calcic lenses at Cima di Gagnone have grossularite components exceeding 75%, but compositions of garnets in rocks transitional between eclogites and metarodingites are less calcic and overlap the range present in the Navajo garnetites.
Temperatures and pressures recorded by Navajo garnetites
Calculations of pressures and temperatures recorded by the garnetite xenoliths depend upon assumptions that mineral grains in contact had been in equilibrium. The ranges of garnet composition within some of the small xenoliths and the sharp compositional boundaries within some grains (Fig. 3) are evidence of disequilibrium, as is the contrast in jadeite component (12 and 25%) between the only two clinopyroxene grains found in garnetite GR1-203 (Table 1). None the less, an approach to equilibrium is consistent with interelement correlations between samples. For instance, in xenolith N379b-GR-4, garnet, chlorite and ilmenite have the most magnesian and chrome-rich compositions within the suite. The systematic Fe–Mg partitioning between garnet and chlorite grains (Fig. 7) is consistent with equilibrium over a temperature range of about 100°C. The partitioning of REE between the single garnet–clinopyroxene pair analyzed for trace elements provides another test. The chondrite-normalized abundances of REE in clinopyroxene form a bell-shaped pattern (Fig. 6c). Both the abundance pattern and the relative depletion of HREE in the clinopyroxene are consistent with equilibration with garnet (e.g. Harte & Kirkley, 1997). Because the P–T and compositional dependences of trace element partitioning between garnet and clinopyroxene are poorly known (Blundy & Wood, 2003), quantitative evaluation of equilibrium using the partitioning is not yet possible for the mineral pair in this xenolith.
Compositions used for temperature calculation were acquired near mutual contacts of chlorite and garnet and of clinopyroxene and garnet. Temperatures were calculated from Fe/Mg partitioning for chlorite–garnet pairs and Fe2+/Mg partitioning for clinopyroxene–garnet pairs at a pressure of 2 GPa (Table 6); no pressures have been calculated from the mineral assemblages in these xenoliths. The thermometer of Krogh (1988) and the revised thermometer of Krogh Ravna (2000) have been applied to garnet–pyroxene pairs; for these applications, ferric iron was calculated in pyroxene from (Na – Al – Cr) in a four-cation formula, and in garnet from (2 – Al – Cr) in an eight-cation formula. Temperatures calculated with the more recent procedure, that of Krogh Ravna (2000), range between 460°C and 600°C and average 530°C. Values calculated with the procedure of Krogh (1988) are 60–90° higher and average about 600°C. The temperatures are not correlated with calculated jadeite or aegirine in pyroxene or with calculated Fe3+/Fe2+ in garnet (Table 6). Moreover, the correlation between calculated Fe3+/Fe2+ in garnet and pyroxene is evidence that the assumptions used to calculate ferric iron have validity. Garnet–chlorite temperatures were calculated by two approaches: that of Grambling (1990) and that of Dickenson & Hewitt (1986) in the modified form described by Laird (1988). All iron was considered ferrous in both garnet and chlorite for these approaches. Both chlorite–garnet methods yield very similar temperatures in the range from 400°C to 500°C, and the average of the five values is about 470°C. The contrast between the lower garnet–chlorite and the higher garnet–pyroxene temperatures may be due to at least two causes other than disequilibrium—assumptions regarding ferrous iron or faulty calibrations of thermometers for these rocks. The garnet–pyroxene temperatures are sensitive to assumptions made in calculating ferric and ferrous iron in such magnesian pyroxenes, as emphasized by Proyer et al. (2004). The garnet–chlorite thermometers were calibrated from garnet–biotite thermometry of metamorphosed pelites in which garnet and chlorite compositions are unlike those in the garnetites. Hence, the calculated temperatures are best regarded only as evidence that the temperatures of garnetite formation were low, probably in the range from 400°C to 600°C. Even if inaccurate, the calculated values are meaningful for comparisons of temperatures calculated by the same methods for eclogite and pyroxenite xenoliths included in the Navajo diatremes.
|
Pressures of garnetite formation are difficult to constrain. Comparisons with metamorphosed rodingites in alpine peridotites are complicated by hydration reactions that occurred as those rocks were exhumed. For instance, Evans et al. (1979)
suggested that metarodingites in an Alpine peridotite had eclogite-facies mineralogies at about 800°C and 2·5 GPa, and that most or all of the amphibole and epidote in those rocks developed during a metamorphic overprint at less than 1 GPa. The absence of amphibole in the garnetites may be a key. Schulze et al. (1987) and Helmstaedt & Schulze (1988) observed that a retrograde assemblage of chlorite–garnet–omphacite formed as a consequence of hydration of some garnet pyroxenites, whereas pargasite–chlorite formed in others. They concluded that the chlorite–garnet–omphacite assemblage formed at pressures greater than about 2·5 GPa, above the stability of amphibole. Neither amphibole nor epidote was identified in the garnetite xenoliths, despite the evidence for the presence of a fluid phase during garnet growth, and hence the chlorite–garnet–clinopyroxene assemblage in these rocks probably also formed near or above about 2·5 GPa.
Relevant xenolith assemblages in the Navajo SUM diatremes
The SUM diatremes in the Navajo volcanic field contain unusual xenolith types that provide context for the interpretation that the garnetites formed in metasomatic reaction zones. These rock types include those peridotites and garnet pyroxenites that contain chlorite and other hydrous minerals and the eclogites. As emphasized by Helmstaedt & Schulze (1988), the diverse lithologies are similar to those in some metamorphosed ophiolite complexes in high-pressure orogenic belts.
Hydrous minerals in spinel peridotite xenoliths include amphibole, chlorite, titanoclinohumite, and antigorite: Smith (1979) concluded that all but antigorite formed by hydration reactions in the mantle, and that at least some of the antigorite may be of similar mantle origin. Peridotite xenoliths also contain chlorite attributed to hydration of mantle at greater depths, below the spinel–garnet transition (Mercier, 1976; Smith, 1995). An unusual rock interpreted by Smith (1995) to record reactions of garnet peridotite with water contains centimeter-scale volumes with contrasting proportions of chlorite, clinopyroxene, orthopyroxene, ilmenite, and titanian chondrodite. Minerals in these volumes appear to have formed in metasomatic reaction zones, consistent with element transport analogous to that during rodingite formation.
Chlorite appears together with a second generation of garnet and clinopyroxene in some garnet pyroxenite xenoliths (Helmstaedt & Schulze, 1979; Benoit & Mercier, 1986; Schulze et al., 1987; Smith, 1995). Chlorite–garnet thermometry records temperatures in the range from 400°C to 500°C for the pairs in apparent textural equilibrium (Smith, 1995), just as recorded by garnet–chlorite pairs in the garnetites (Fig. 7). Garnets formed with chlorite in the pyroxenites have compositions near Gr28Prp36Alm36, similar to those in some garnetites (Fig. 2), as also noted by Helmstaedt & Schulze (1988). However, clinopyroxene in the retrograde assemblages of the garnet pyroxenites has 37–47% Jd component, more sodic than the range from 3% to 25% found for pyroxene in the garnetites.
Eclogite xenoliths record garnet–pyroxene temperatures between about 500°C and 700°C (e.g. Helmstaedt & Schulze, 1988; Smith et al., 2004), similar to the values calculated for the garnetites (Table 6). Coesite has been identified in one eclogite, establishing a minimum pressure of about 2·5 GPa (Usui et al., 2003). Three eclogites record pressures in the range 2·6–3·4 GPa by garnet–phengite–omphacite barometry, and the calculated pressures and temperatures are consistent with the presence of lawsonite in these rocks (Smith et al., 2004). Textures of the eclogites indicate the presence of a water-rich fluid during recrystallization (Smith & Zientek, 1979), and the formation of sodic eclogite and associated jadeite and omphacite pyroxenite has been attributed to hydrous metasomatism (e.g. Helmstaedt & Schulze, 1988; Wendlandt et al., 1993).
|
GENESIS OF THE GARNETITES |
|---|
The garnetites are unlikely to be igneous in origin. First, they record only subsolidus conditions. Chlorite, all of which appears primary, occurs in five of the xenoliths, and these rocks are otherwise typical of the suite. Temperatures calculated from garnet–chlorite and garnet–clinopyroxene pairs by all methods are in the range 400–700°C, and those calculated by the more recent methods are in the range 400–600°C (Table 6). Many garnets contain compositional gradients that are sharp on a scale of several micrometers, as documented by the back-scattered electron images in Fig. 3 and by the traverses plotted in Fig. 5. Such compositional gradients would have at least partly annealed, if cooling had been slow enough to reset the garnet–clinopyroxene and garnet–chlorite temperatures. Therefore, the mineralogy, geothermometry, and textures are metamorphic. Second, although the temperature constraints could be consistent with low-temperature metamorphism of an igneous protolith, the xenolith compositions are unlike those of the rare garnetites attributed to igneous processes. The igneous garnetites of mantle origin described by Exley et al. (1983)
and Zhang et al. (2004) are more magnesian than the xenoliths. Those interpreted as deep-crustal restites by Rivalenti et al. (1997) contain less calcic garnet. Thus there is no evidence that the unusual bulk compositions of the Navajo garnetites formed by an igneous process.
The bulk compositions, mineralogies, and textures of the garnetites instead are attributed to formation in metasomatic reaction zones at contacts of mafic rock and peridotite within the continental mantle lithosphere. The reactions are inferred to be a consequence of fluid flow from hydrating peridotite into the mafic rock, a process analogous to that which forms rodingite. Rocks that are monomineralic, or nearly so, are commonly formed in metasomatic zones (Thompson, 1959). Coleman (1967) and many others have described rodingites that formed by metasomatism at contacts of serpentinized peridotite. The fracture-related garnet growth establishes that a hydrous fluid was present during garnetite formation, as required for the process. Although the garnets are unlike those of rodingites, the differences may be due to relatively higher temperatures and pressures of garnetite formation. REE patterns of the garnets are also consistent with formation in metasomatic reaction zones. The patterns are not like those for many mantle garnets, in which ‘convex-up’ chondrite-normalized plots decrease smoothly from Lu to La. Rather, in four of the five rocks, abundances of the middle REE (MREE) are relatively high, more similar to the ‘sinusoidal’ or ‘humped’ patterns interpreted as consequences of metasomatism by Hoal et al. (1994), Roden & Shimizu (2000), Burgess & Harte (2004), and Zhang et al. (2004).
Garnetite formation during serpentinization is consistent with the stability of antigorite in the mantle. Aluminous antigorite is stable to temperatures above 660°C at 2 GPa (Bromiley & Pawley, 2003), hotter than almost all temperatures calculated for the garnetites (Table 6). Bromiley & Pawley (2003) also demonstrated that antigorite with 3·1 wt % Al2O3 is stabilized to higher temperatures than is pure antigorite, and antigorite in Navajo peridotite xenoliths contains as much as 3·9 wt % Al2O3 and 1·2 wt % Cr2O3 (Smith, 1979).
Compositions of the minor minerals are consistent with the hypothesis that peridotite was involved in the metasomatic process. Clinopyroxene has calculated Fe2+/(Fe2+ + Mg) in the range 0·05–0·08. Chlorite and phlogopite have Fe/(Fe + Mg) of 0·05–0·08 and 0·10, respectively. These ratios are similar to those in some peridotites. The minor minerals also have appropriate Ni contents. Analyzed chlorite has NiO in the range 0·24–0·44 wt %, with an average of 0·34 wt %. For comparison, chlorite in equilibrium with olivine in alpine peridotites has NiO in the range 0·19–0·23 wt % in the rocks studied by Trommsdorff & Evans (1969, 1974) and Smith (1979).
The ranges of mineral compositions and the varieties of compositional zoning within garnet may record sample positions within individual reaction zones and the evolution of the zones with time. MgO and Cr2O3 are zoned to higher values at rims of some garnets (Figs 4a and 5b), consistent with changes expected as a mafic lithology is infiltrated by water involved in peridotite hydration. Other features of the zonations are less easy to interpret. For instance, CaO is relatively low in late-stage garnet in one xenolith (Fig. 5a) and relatively high in another (Fig. 5b). The ranges in trace element composition may be due to the presence of gradients of chemical potential at a variety of scales, as discussed for metarodingite formation by Frost (1975): Zr concentrations in garnet (Table 2) range from a low of 2 ppm in the rock with relatively abundant zircon (GR1-201) to a high of 17 ppm in a rock in which zircon was not found (ATG-GRP1-B).
Rodingite formation commonly is interpreted as a process that occurs in the upper crust (e.g. Evans, 1977), but the ages deduced from U–Pb and Hf isotopic data instead are compatible with geochronology of the Plateau mantle. U–Pb ages establish that much of zircon in garnetite GR1-201 formed from 85 to 60 Ma, but no significant Phanerozoic thermal event has been recognized in geochronological studies of crustal xenoliths in the Navajo diatremes (Condie et al., 1999; Selverstone et al., 1999; Crowley et al., 2004). Most of the age range recorded by concordant garnetite zircon is within the 81–33 Ma period of growth of concordant zircon in the associated eclogite xenoliths (Usui et al., 2003; Smith et al., 2004), and eclogite recrystallization has been attributed to mantle processes, either in the subducted Farallon slab (Usui et al., 2003) or in the overlying mantle wedge (Wendlandt et al., 1996; Smith et al., 2004). Garnetite genesis within the Farallon slab is precluded by the U–Pb and Hf isotope data that establish inheritance of Proterozoic zircon, but the data are consistent with garnetite formation in the Proterozoic mantle wedge during Farallon subduction.
|
EVOLUTION OF THE MANTLE BELOW THE DIATREME |
|---|
Age and stability of lithosphere
The inherited zircon in garnetite GR1-201 adds an important constraint for the evolution of the lithosphere of the Colorado Plateau. The eclogites with the most basalt-like compositions have Nd model ages of 2·7 Ga (Roden et al., 1990
) and 1·5–1·8 Ga (Wendlandt et al., 1993), but interpretation of these ages is hindered by the REE metasomatism that affected the suite. Re-depletion Os model ages of peridotite xenoliths from a minette plug in the Navajo field range from 1·1 to 1·8 Ga; the range may be due to addition or loss of Re before eruption (Lee et al., 2001). The Hf depleted mantle model age of at least 1·8 Ga for garnetite zircon is a more robust indicator of when the Plateau mantle was stabilized. That age is matched by the oldest Hf depleted mantle model ages of zircon cores in xenoliths from the lower crust, also 1·8 Ga (Crowley et al., 2004), and it is consistent with Nd model ages of crustal xenoliths that establish the main crust-forming event at about 1·85 Ga (Wendlandt et al., 1993). The correspondence of ages for mantle and crustal formation is evidence that the crust and uppermost mantle have been coupled together since initial crustal formation.
Thermal histories and a tectonic model
Calculated temperature histories provide insights into the depth of garnetite formation. A possible pre-Laramide geotherm (Fig. 12) was calculated for a lithosphere of 200 km thickness that has a basal temperature of 1300°C and a surface heat flow of 56 mW/m2; the heat flow is the average of the two values closest to the diatreme in the compilation of Minier & Reiter (1991). Geotherms also are plotted for a model of the lithosphere during flat subduction of the Farallon slab, 700 km from the trench with a relative plate convergence of 10 cm/year. The geotherms for the period during Farallon subduction were calculated using a standard finite difference conductive thermal model similar to but simpler than that used by Spencer (1996). He calculated geotherms for models in which the Farallon slab sheared off and replaced the lower part of the Plateau lithosphere: the most significant difference in the model used here is that the geotherm in the Farallon plate at the trench was not adjusted for the age of the subducting slab. The interplate contact was assumed to be at 150 km: this thickness for the remnant lithosphere is consistent with the minimum value of 120 km based on analyses of Os in xenoliths erupted at about 25 Ma (Lee et al., 2001) and with the value of 120–150 km based on interpretations of data from the LA RISTRA seismic line (West et al., 2004). The LA RISTRA array passes about 5 km from the diatreme. The crustal thickness of about 47 km is constrained by receiver function analysis of data from that same seismic array (Wilson et al., 2003). The model used for geotherm calculation is simplified and the necessary assumptions are not well constrained, but one conclusion is relatively robust: before and near th